Qualify IPv6 link-local addresses with scope_id

[ghost]

When a service is advertised on an IPv6 address where
the scope is link local, i.e. fe80::/64 (see RFC 4007)
the resolved IPv6 address must be extended with the
scope_id that identifies through the "%" symbol the
local interface to be used when routing to that address.
A new API is provided to return qualified addresses
to avoid breaking compatibility on the existing
parsed_addresses().

This patch:

• Extracts the scope_id from an IPv6 link-local address
  and return it in the retrieved information in ServiceInfo.
• Adjusts the browser test to be able to show IPv6 link-local
  addresses by using the new method parsed_scoped_addresses().
• Uses scope_id in unicast replies to IPv6 link-local addresses,
  and adds some unit tests to verify the new interface provided.

Co-authored-by: Lokesh Prajapati lokesh.prajapati@ncipher.com

Closes #314

[coveralls]

Coverage decreased (-0.9%) to 92.446% when pulling 620564f on nCipherSecurity:master into 4da1612 on jstasiak:master.

[vazhnov]

I see an error in automatic Travis build with Python 3.5 only:

E     File "/home/travis/build/jstasiak/python-zeroconf/zeroconf/__init__.py", line 1899
E       return [f"{addr}%{self.interface_index}" for addr in ll_addrs] + other_addrs
E                                              ^
E   SyntaxError: invalid syntax

[jstasiak]

Thank you! I'll need to think about this, I'm not a fan of a separate API for this (parsed_scoped_addresses(), scoped_id being separate from address variables etc.) but we need to support this in some way.

[ghost]

Is there a way to run locally coveralls to check the coverage reports? I would like to understand if there is anyway to improve coverage in the proposed patch as well.

[jstasiak]

You can use our test_coverage make target to use the standard coverage module and print the output.

[yoyko]

I ran into a problem with this implementation: when a device reports both ipv4 and ipv6 addresses over ipv4 (well, usually it reports both addresses over both ipv4 and ipv6 ;) and python-zeroconf listens only on ipv4 (which is the default), then scope_id won't be added:

addr: 10.147.0.20, port: 5353, v6: []
+ TestService._test._tcp.local.: ['10.147.0.20', 'fe80::2086:3656:3eb9:4d45']

Additionally, if the device reports both addresses, but only over ipv4 (yeah, I know, this is a weird behavior and could be considered a bug...), setting python-zeroconf to listen on all actually makes it worse, becase with dualstack (IPV6_V6ONLY==false), ipv4-in-ipv6 addresses from socket.recvfrom have scope-id set to 0, which then gets set on the ipv6 address:

addr: ::ffff:10.147.0.20, port: 5353, v6: [0, 0]
+ Test._test._tcp.local.: ['fe80::2086:3656:3eb9:4d45%0', '10.147.0.20']

Test program:

from zeroconf import Zeroconf, ServiceBrowser, IPVersion
import sys
import time

class L:
    def add_service(self, z, t, n):
        info = z.get_service_info(t, n)
        print(f"+ {n}: {info.parsed_scoped_addresses(IPVersion.All)}")

ip_version = None
if len(sys.argv) > 2 and sys.argv[2] == "all":
    ip_version = IPVersion.All
z = Zeroconf(ip_version=ip_version)
l = L()
b = ServiceBrowser(z, sys.argv[1], l)

time.sleep(2)

The first message is from:

diff --git a/zeroconf/__init__.py b/zeroconf/__init__.py
index c633d5f..b074533 100644
--- a/zeroconf/__init__.py
+++ b/zeroconf/__init__.py
@@ -1410,6 +1410,7 @@ class Listener(QuietLogger):
             self.log_exception_warning('Error reading from socket %d', socket_.fileno())
             return
 
+        print(f"addr: {addr}, port: {port}, v6: {_v6}")
         if _v6:
             log.debug('IPv6 scope_id %d associated to the receiving interface', _v6[1])
 

The first test was run against

avahi-publish-service  TestService _test._tcp 1234

[bdraco]

Closing this since the OP is now a ghost.

[ibygrave]

Could this be reopened? The OP was working with me, and I have taken over this issue. I've opened a new PR for it.

[bdraco]

New PR is #343